Praat, J.P.
Maber,J.
Yule, I.J.

Initial hopes were high for the adoption of precision agricultural systems for site specific management in mainstream agriculture. The system automatically records and graphically highlights areas on farms where efficiencies can be improved. However, adoption of precision agricultural systems appears to have faltered (Lowenberg-DeBoer, 1999). The problem is that it is unclear if farmers will achieve sufficient return from their investment in the technology. This report relates experience with the technology to date in New Zealand, outlines barriers to further uptake of the technology and proposes strategies for integrating the technology into existing operations.

New Zealand Experience

While the agricultural aviation industry in New Zealand has used Global Positioning Satellites (GPS) for swath guidance for over 10 years, few operators have used this technology for ground based vehicles (e.g. combines, tractors, sprayers). Several operators, spread throughout New Zealand, have yield mapping equipment on combines used for cereal harvesting. Two, one in the South Island and one in the North Island have been the subject of some detailed study. These studies indicate similar trends to overseas studies in variation of soil nutrient levels (Table 1). Interestingly the values calculated for both studies show similar variation and although the variability is clear the relationship with yield is less so. A number of studies around the world have indicated higher levels of P and K on lower yielding areas. Mohamed et al (1997) suggested this was due to over-application through the spreading of blanket fertiliser rates with insufficient turnover of nutrients in lower yielding areas. This suggests that savings could be made on P and K applications using variable rate fertiliser application while maintaining an acceptable level of fertility.

¥ - 260 soil samples taken from 46 ha from three fields in Canerbury

The system under study in the North Island is operated by John Austin Ltd. Yield maps have been produced for three seasons for approximately 300 fields with the John Deere GreenStar™ system. These maps have been used in the field. A number of yield reducing factors were identified, and a hierarchy of factors soon became apparent. The most easily recognised factors included hybrid, drainage, weeds, insects, old field boundaries and limitations due to soil type. The costs and benefits associated with these limitations and potential remedies can be confidently costed. For example Figure 1 shows that approximately 6.5t/ha yield reduction was due to wet soil conditions which cost NZ$1527/ha (@NZ$235/t maize). The capital required to improve drainage can be balanced against a quantifiable increase in production.

Figure 1 Areas of poor yield resulting from wet soil conditions

 
 

In another field a superior hybrid was identified and a portion of the same field with light soil was retired (planted in trees) as the yield was not sufficient to cover production costs. The relationship between soil nutrient level and yield was less obvious although one field in particular appeared to benefit from site specific soil sampling. A field was soil sampled on a grid basis (1 sample/ha) to test links with yield variation. Potassium levels appeared to be associated with yield variations. Areas of low and high K were identified and an application map was prescribed using three rates of muriate of potash (0, 50 and 100 kg/ha). The subsequent yield map was more even as previously low yielding areas yielded close to the field average. The cost of intensive grid soil sampling appeared to be justified in this case. However, this tended to be the exception rather than the rule. Soil quality was another factor which was harder to identify as a limitation to yield. For example identification of layers in the soil which may restrict root penetration or areas where soil structure (density, stability, porosity) is limiting growth is often difficult due to temporal and moisture changes.

In the Canterbury study soil depth was identified as having the strongest correlation to yield on the downland site. Two further sites on the Canterbury plains have greater variability in soil depth despite being much flatter. These sites were irrigated which if scheduled correctly should nullify the effects of soil depth. The study is just beginning the second year and caution should be exercised in interpreting these results as the conditions at the time were exceptionally dry, McBratney (1999) observed that temporal variation can be greater than spatial. The three year study is still in the initial phase, having collected two years of yield maps and extensive field data from selected sites on the three properties for one growing season.

Factors limiting the uptake of precision agriculture systems

The economics of swath guidance are easily identified. Improvements such as 5 to 8 % increase in accuracy of bout width, the ability to spray at night and improved confidence in novice operators are measurable economic benefits swath guidance technology (Strautman, 1999). Minimising overlaps and misses optimise labour, fuel and agrichemical inputs. The profitability of site specific management are however less certain. Initial yield mapping systems have been developed on headers for two main reasons. Firstly; a larger market than any other harvesting machinery and secondly, it is simpler to accurately measure the flow of a granular material. However it must be remembered that most of the crops going through the header are less valuable per hectare than many other crops such as potatoes, vegetables and wine grapes. Our initial focus has been on savings of input costs but this is less profitable than increasing yield and it is that we should be concentrating on. For the farmer the aim of increased profitability is clear, the mechanism of how to achieve it is not. Many of the support structures required are not yet in place.

At this stage it is likely that only large farmers or contractors will be able to justify the $20,000 to $50,000 investment required to implement precision agriculture systems as they are able to spread the fixed cost over a large area. In the case of the large farmer, value is required from improved efficiencies of inputs. In the case of the contractor, value may be obtained from either attracting more clientele, charging a premium or, which is more likely the case, maintain market share. Part of the drive for adopting these systems is related to the perceived requirement of future markets to require rigorous audit trails in order that "good agricultural practice" is evidenced. Those involved in the export trade have already witnessed this as part of their customer’s quality assurance schemes.

Opportunities

What is missing? Why have farmers not embraced the technology with both arms? As researchers we constantly hear farmers (decision makers) complaining that the system does not tell them what they need (or want) to know in order that they can manage their production unit more efficiently. The critical aspects or the "need to know" parts of the business include the relationships between inputs, outputs and externalities. The data on these aspects must have appropriate precision and be available in a timely manner. The production unit, the manager and the environment are described in Figure 2 as a typical pressure-state-response model (Smyth and Dumanski, 1994). The pressure side of the model is the issues which may impact on the current position (state) which the decision maker has little or no control over. The state means the current position that the farmer is in. The response is action taken to deal with these issues.

Figure 2 Description of an agricultural production unit
 

Pressure		State		Response
Prices	«	Product mix	¬	Decision maker
Market Signals		Yield
Weather		Risk strategies

 
 
 

The quality of the response is determined by a number of things but the response is generally made with imperfect information or data eg. price and weather predictions. Precision agriculture incorporates the computing, digital and electronic technologies of GIS/GPS, DSS, monitoring and control equipment. The combination of these technologies should give farmers superior information to that they have had in the past in order to make more informed decisions. The system would include information such as; expected prices, insect and weed threshold levels, weather and crop development data. This must be presented in a form which the decision maker can use easily and have confidence in. We need to move on from describing relationships between yield and various crop/soil factors for individual fields to develop decision support systems which can help farmers make sense of yield maps on their farm as well as monitoring temporal factors that will have a direct impact on yield and yield spatial distribution.

Reading the scientific literature and conference proceedings indicates that hundreds of studies have been completed over the last ten years. Many of these studies have attempted to describe the relationship between yield and factors limiting yield. Almost invariablely they show high levels of spatial variation in soil properties and spatial variation in yield. Some identify the limiting factor for yield in that particular case. Longer term trials often come up with the idea of identifying stability in the system. For example Blackmore (1997) examined the consistency of yield between years in order to place areas under three management classes. These were; high yield and stable, low yield and stable and unstable. The proportions of the total paddock area placed in each category and will clearly vary between sites as will the effect of crop management and other temporal factors. This approach works when considering a stable system and we can differentiate areas of high and low yield, but what happens if part of the system is not stable. In this case we have to rely on shorter feedback loops and more monitoring through the crop’s growth cycle in order to make management decisions.

This perhaps begins to give us a clue as to the apparent lack of adoption of precision farming. What we have done is to try to convince farmers to invest in a concept that:

1. Requires a major capital investment and shift in the way the business is managed.
2. Researchers have a lack of certainty as to the benefits. Most are careful to use the words "Potential Benefits" as they recognise no two cases are alike.
3. Requires a whole raft of skills that farmers do not feel they have in order to support it.
4. Has been described as having long term "Potential Benefits" in an industry where most are pressurised into making decisions for short term survival.

There is a wide gap between the work researchers have carried out and what needs to be done in order to facilitate the uptake of precision farming methods. We still seem to be in the experimental phase but we need to move to the developmental. This prolonged experimental phase is perhaps indicative of our own uncertainty as we recognise that most cases are different. Having done that we should move to examining ways the technology could be more readily adopted into the management structure of agricultural and horticultural businesses. There should be greater focus on the development of management tools that will allow the farmer to complete complex analysis tasks without having to be a software writer. These tools must be reliable, flexible, easy to use and at a price farmers can afford. They must be able to integrate and communicate with other computer systems and offer financial analysis as well as physical.

Appropriate DSS are required. Farmers need planning software to determine economic/environmental trade-offs on farms. It is unlikely that a DSS will exist for every farmer and every situation he strikes every year. Inevitably DSS models have to be simplified to get them working. Generic models which provide return on investment type information and the impact of decisions on the whole farm are required. These models must be easy to customise using the farmers own data so that as the farmers collects his own data from year to year this data can be added to improve the accuracy of the model. (Rosiland Buick pers com. 1999). Developing such a will be difficult but without such a focus much software will miss it’s mark.

There are a lot of data available now which are under-utilised. An example of this in New Zealand is that of fruit harvesting. Currently apples are picked into bins with the block, variety and picker noted on card stapled to the bin. When the bin gets to the grader the card is removed and the link to packout yield and the orchard operations is lost. What is needed is tracking system for individual bins which is compatible with the grading operation and can accommodate orchard data such as the tree or group of trees the fruit came from, the management of those trees (pruning, irrigation, fertiliser) and the associated costs of production. Work should focus on making that data available to the decision maker in a form they can use. Precision agricultural systems should be thought of as management tools rather than research tools.

Spray drift is another good example. A great deal is known about the various factors that each have an effect e.g. droplet size, wind speed, emission point. What is missing is a system that brings this information together in a way that a sprayer operator can use. A current project aims to use precision agriculture systems to provide a link between what we know about spray drift from orchards, the weather at the time of spraying and the features of the orchard and it’s surrounds. The aim is to provide a system to help the sprayer operator manage the risk of spray drift. Potential spray drift will be predicted using available models based on realtime data on wind speed and direction. The resulting spray concentration levels will be overlaid with threshold levels set for the properties surrounding the orchard. From this a "safe spray" area may be identified to the operator.

Horses for courses.

For those contemplating investment in precision agricultural system our advice would be to get into it. Yield mapping is the first logical step of this process and we believe there will be benefits from this although the nature of these benefits may be unclear initially they will be surprising. It is possible to produce yield maps for combineable and root crops now. These can be used in their simplest form by using your feet and walking across fields with map in hand and observing the crop and factors such as weeds, insects, drainage, soil quality cultivation / planting effects. Farmers should not be put off by thinking that variable rate technology (VRT) and associated sampling / scouting costs are required to extract a viable return from precision agriculture. They may miss the opportunity to gather some very cost effective data in the form of yield contour maps for their fields. Generally three to five years of yield maps for a field are required before it will be obvious how inputs such as agrichemicals, cultivation, fertiliser and seed should be economically adjusted across a field. These decisions can and should be made at a later date. Even three years of yield maps may not be enough as two of the three may be unusually dry or wet although extreme years may in fact help to segregate discrete management units in a field.

Conclusion

Until recently the author was immersed in a business which used precision agricultural systems to provide yield maps for clients and for a cropping operation. After three years of experience the owner of that business is unsure as to what his next step should be in terms of further investment in the technology. The benefits of the system were there and observed in terms of identifying and quantifying some obvious limitations to yield. However, there is a feeling that the system has not quite delivered on expectations. The next step should be focused on providing the information the decision maker needs to know in order that he can maximise his profits. This is important to the farmer and should be the focus of the workers in this area.

References

Blackmore, B.S. Larscheid, G. (1997) Strategies for managing variability. In Precision Agriculture 97: Proceedings of the 1^st European Conference on Precision Agriculture, J.V. Stafford (ed.), Oxford, UK:Bios. pp.851-859.

McBratney, A.B. Whelan,B.M. (1999) The ‘Null Hypothesis’of precision agriculture. In Precision Farming 99: Proceedings of the 2^nd European Conference on Precision Agriculture, J.V. Stafford (ed.) Sheffield Scientific Press. pp.947-957

Mohamed, S.B. Evans,E.J. Shiel,R.S. (1997) Mapping techniques and intensity of soil sampling for precision farming. In Precision Agriculture 97: Proceedings of the 1^st European Conference on Precision Agriculture, J.V. Stafford (ed.), Oxford, UK:Bios.

Smyth, A.J. and Dumanski, J., 1994: Progress towards an international framework for evaluating sustainable land management (FESLM). 15^th World Congress of Soil Science, Acapulco, Mexico July 1994: Transactions 6a:373, Symposium ID-8.

Strautman, B., 1999: Parallel swathing saves time, too. Western Producer Farming, Vol 2, 6:5-10 June-July 1999.

Lowenberg-DeBoer, J. 1999: Precision ag adoption rates plateau at 25%. @g News http://www.agriculture.com/scgi/AgNews, 16 August, 1999.

Personel Communications

Dr. Rosiland Buick, Trimble Navigation New Zealand Ltd. 1999